miknas-compoundos（米克纳斯化合物）

CompoundOS - AI Operating System Implementation

Core Concept

CompoundOS is a self-improving AI Operating System that eliminates "context reset" - where scattered AI tools create disconnected data and lost context. The system compounds intelligence daily through a learning loop.

Key benefits:

• Self-improving: Every task makes the system smarter
• Anything Tool: AI builds tools/workflows instead of buying SaaS
• Frictionless: Eliminates bottlenecks, enables systematic high-leverage work

Quick Start: 3-Step Implementation

Step 1: Define Strategic Layer (Component 1)

Create master document with these elements:

Required fields:

• Big Obsessional Goal (BOG): Your single, driving ambition
• Current Bottleneck: The #1 thing blocking progress
• Target Audience: Who you serve and their pains
• Positioning: How you're uniquely positioned to win

See assets/strategy-template.md for template.

Step 2: Create Agent with Strategy

Feed strategic document into AI agent's permanent instructions. This ensures:

• Every decision is filtered through the strategy
• Agent can push back on misaligned requests
• Context is maintained across sessions

Step 3: Enforce Filter

Always prompt AI as "Chief of Staff":

1. Review strategic document before executing
2. Score tasks against business objectives
3. Surface ONE needle-moving action daily

Implementation Workflow

Phase 1: Foundation (Components 1-3)

1. Strategic Layer - Define core (see above)
2. Prioritization Engine - Set up daily review cadence

- Review backlog against strategy - Score tasks on strategic alignment - Output: ONE action to execute today

1. Knowledge Management - Set up memory system

- Capture insights, decisions, outcomes - Auto-categorize by department/project - Enable retrieval before new tasks

See references/knowledge-setup.md for detailed implementation.

Phase 2: Execution Layer (Components 4-6)

1. Central Ops - Build workflow automation

- Document SOPs for repeatable processes - Create automated task pipelines - Establish reproducible processes

1. Department Agents - Deploy ACRA agents

- See references/department-agents.md for agent templates - Each agent holds only department-relevant context - Specialized capabilities per department

1. Projects - Set up cross-functional orchestration

- Shared context when goals span departments - Example: Product launch = Attract + Deliver collaboration

Phase 3: Learning Layer (Components 7-9)

1. Auto-Capture - Enable self-improvement

- Log all decisions, actions, outcomes - Feed data into knowledge system - See references/learning-loop.md

1. Communication Layer - Set up data gateways

- Human-to-Machine: Voice, text, structured input - Machine-to-Machine: APIs, CRMs, webhooks

1. Metrics & Monitoring - Establish operating rhythm

- See references/metrics-cadence.md - 5 cadences: Daily, Weekly, Monthly, Quarterly, Annually - Performance signals feed back to Strategic Layer

ACRA Framework Quick Reference

Department agents follow ACRA structure:

Support functions: Finance, HR, Legal (as needed)

See references/department-prompts.md for agent prompt templates.

The Compounding Cycle

Strategic Layer → Prioritization → Execution (Ops/Departments/Projects)
         ↓
   Auto-Capture
         ↓
┌────────────────────┴────────────────────┐
↓                                          ↓
Knowledge Management                  Metrics System
↓                                          ↓
└───────────────→ Learning Loop ←────────┘
                      ↓
         Updates & Refines Strategy

Result: Your AI wakes up smarter each day.

Component Interdependencies

• Strategic Layer → Guides Prioritization Engine (Component 2)
• Auto-Capture → Feeds Knowledge Management (Component 3)
• Department Agents → Use Central Ops for workflows (Components 4-5)
• Metrics System → Sends signals to Strategic Layer (Components 1-9)
• Communication Layer → Connects all components (Component 8)

Common Patterns

Daily Operations Pattern

1. Morning: Prioritization Engine surfaces ONE needle-moving action
2. Mid-day: Department agents execute specialized work
3. Evening: Auto-Capture logs outcomes, Metrics reviews performance
4. Night: Learning Loop updates knowledge, refines strategy

New Task Pattern

1. Input: Request enters via Communication Layer
2. Filter: Prioritization Engine scores against strategy
3. Route: Task assigned to appropriate department agent
4. Execute: Agent completes work with Central Ops support
5. Capture: Auto-Capture logs entire process and outcome
6. Learn: Knowledge Management extracts insights

Project Launch Pattern

1. Define: Project scope shared across relevant departments
2. Coordinate: Cross-functional agents establish shared context
3. Execute: Each department contributes specialized work
4. Monitor: Metrics System tracks project KPIs
5. Review: Post-mortem captured, lessons learned

Troubleshooting

Context Disconnect

Symptom: AI forgets previous decisions or context

Solution:

• Ensure Auto-Capture is logging everything
• Check Knowledge Management retrieval is working
• Verify Strategic Layer is being applied as filter

Analysis Paralysis

Symptom: Too many priorities, can't decide what to do

Solution:

• Strengthen Prioritization Engine scoring
• Limit to ONE needle-moving action per day
• Revisit Strategic Layer for clarity

Department Silos

Symptom: Teams not sharing context, duplicated work

Solution:

• Use Projects for cross-functional goals
• Ensure shared context is orchestrated
• Check Communication Layer integrations

No Learning Occurring

Symptom: System not getting smarter over time

Solution:

• Verify Auto-Capture is active
• Check Knowledge Management is extracting insights
• Ensure Metrics feedback loop is reaching Strategic Layer

Best Practices

1. Start small: Implement Components 1-3 first, then expand
2. Define before build: Strategic Layer must be solid first
3. Capture everything: Auto-Capture is non-negotiable
4. One action per day: Prioritization Engine enforces focus
5. Review regularly: Metrics cadence must be maintained
6. Iterate strategy: Learning Loop must update Strategic Layer

Reference Materials

When to Use This Skill

Use CompoundOS when:

• Building AI-powered business operations systems
• Implementing agentic workflows with departmental specialization
• Creating self-improving business intelligence systems
• Eliminating context reset across multiple AI tools
• Establishing compounding intelligence architectures
• Setting up automated task prioritization and execution
• Designing cross-functional AI agent teams

*CompoundOS: Your business intelligence compounds daily.*